Two-Parameter Ansatz for the Violation of Eigenstate Thermalization
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.
Kai: We're looking at the paper: "Two-Parameter Ansatz for the Violation of Eigenstate Thermalization".
Paper summary: Kai: So we're looking at this paper called "Two-Parameter Ansatz for the Violation of Eigenstate Thermalization." Basically, the authors are trying to create a single framework to describe how ETH can break down, moving beyond just conventional ETH and its total collapse.
Mira: That’s right. The main idea is introducing a two-parameter ansatz that unifies fading ergodicity and trapped ergodicity <ref:2608.02744#pg4>. They test this framework using spin-one-half J1-J2 chains with some on site disorder and a linear potential, and they find that what they see is consistent with trapped ergodicity <ref:2608.02744#pg4>.
Lev: So if I understand this right, they're using this ansatz to categorize the different ways ETH can fail in a system, separating fading ergodicity and trapped ergodicity based on how they scale with the system size L <ref:2608.02744#pg3>.
Kai: Exactly. Conventional ETH is described by Srednicki's ansatz for matrix elements, where the variance of those diagonal elements scales like exp-L minus xi over eta zero <ref:2608.02744#pg7>. But this paper introduces a new way to look at that scaling using ξ(h, L) = theta of h times L plus lambda of h <ref:2608.02744#pg9>.
Mira: And plugging that into the scaling relation gives us an expminus one minus theta of h times L minus lambda over eta zero scaling <ref:2608.02744#pg9>. That means the limit theta equals zero is conventional ETH, which we know well >
Kai: Right. So we have two limits for this new framework: either theta goes to one, which they call fading ergodicity, or lambda goes to O of L, which they call trapped ergodicity <ref:2608.02744#pg3>.
Lev: From an engineering standpoint, the distinction between those two limits is what matters when you actually try to build hardware. Fading ergodicity suggests a different kind of behavior where theta increases with h <ref:2608.02744#pg8>.
Mira: But trapped ergodicity, on the other hand, is characterized by lambda equals ξ(h), which increases with h and shows a shift on the horizontal axis of the decay of sigma squared consistent with conventional ETH <ref:2608.02744#pg3>.
Kai: The numerical evidence in those spin chains is interesting because for weak disorder, the decay of sigma squared looks like conventional ETH, but at larger disorder, you see these horizontal shifts that point toward trapped ergodicity <ref:2608.02744#pg4>.
Lev: And what they found regarding the parameters theta of h and lambda of h is that for small h, they are roughly constant near zero, but for trapped ergodicity, lambda of h is an increasing function of h <ref:2608.02744#pg4>. That’s a clear signature that differentiates it from fading ergodicity where theta increases with h >
Mira: The paper suggests that the scaling variable L minus lambda of infinity has perfect scaling collapse in the regime when the logarithm of sigma squared scales like L minus lambda of infinity <ref:2608.02744#pg4>. This really supports trapped ergodicity as a relevant description for ETH violation >
Kai: So, to summarize the main point of "Two-Parameter Ansatz for the Violation of Eigenstate Thermalization," it’s introducing this unified way to describe the continuous onset of ETH violation between conventional ETH and its complete breakdown <ref:2608.02744#pg1>.
Mira: The authors show that trapped ergodicity applies to widely studied models like the disordered J1-J2 spin-one-half chain, which is significant because it provides a common theoretical framework for these kinds of finite system effects >
Lev: It opens up new perspectives on how quantum thermalization connects with ergodicity breaking transitions in finite systems <ref:2608.02744#pg1>.
Kai: The authors also point out their limitation, which is that they focus on testing this framework specifically in spin-one-half J1-J2 chains with on site disorder and linear potential <ref:2608.02744#pg1>. They show what happens in those specific models >
Mira: And the conclusion is really about providing a common theoretical structure for describing these two distinct scenarios of ETH breakdown, showing that trapped ergodicity is a key mechanism here >
Kai: So, to wrap up, this paper gives us a tool—a two-parameter ansatz—to distinguish between fading ergodicity and trapped ergodicity when it comes to ETH violation <ref:2608.02744#pg1>.
Mira: It helps us understand that for certain systems, like the ones they tested, trapped ergodicity is what we actually see happening at the boundary of ergodicity >
Lev: So, if you're thinking about running this on real hardware, understanding which scenario applies—fading or trapped—will dictate how much overhead you need to account for in your error correction <ref:2608.02744#pg1>.
Kai: That’s the connection there. We see the math, but it translates directly into practical considerations about fidelity and complexity >
Conclusion: Kai: So, this paper is about proposing a way to describe when Eigenstate Thermalization starts to break down in quantum systems, and they call it the Two-Parameter Ansatz for the Violation of Eigenstate Thermalization.
Mira: Yeah, basically they’re trying to unify two different ways this thermalization breaks down—fading ergodicity and trapped ergodicity—into one mathematical structure <ref:2608.02744#pg4>.
Kai: It’s really about giving us a single language to talk about these distinct scenarios, moving past just the standard ETH picture <ref:2608.02744#pg1>.
Lev: So if I get this right, they’re using this two-parameter setup to distinguish between fading ergodicity and trapped ergodicity based on how the system size L affects the variance of those diagonal matrix elements <ref:2608.02744#pg9>.
Mira: Exactly. They show that in systems like spin chains with some disorder, what they observe fits this framework, specifically pointing towards trapped ergodicity as a key mechanism <ref:2608.02744#pg3>.
Kai: It’s interesting because it suggests that the behavior we see in these models is actually linked to finite-size effects right at the edge of ergodicity, rather than just one simple collapse into total chaos.
Mira: Right. The authors use numerical evidence on those spin chains to show that trapped ergodicity shows up when the horizontal shifts in their measurements match the scaling they predicted for trapped ergodicity <ref:2608.02744#pg4>.
Lev: From my side, what this means is that if we’re building real quantum hardware, understanding which one of these two scenarios applies—fading or trapped—will dictate the kind of error correction overhead we actually need to account for <ref:2608.02744#pg1>.
Kai: So they’re laying out a common theoretical structure that links quantum thermalization directly to ergodicity breaking transitions, which opens up new avenues for how we look at finite systems and their properties.
Mira: It really provides a clearer roadmap for where the physics is going when we study these kinds of complex interacting quantum systems <ref:2608.02744#pg1>.
Lev: Next time, I think we should talk about how this kind of scaling applies to those real, noisy devices we’re actually trying to prototype.
Kohei Ohgane, Lev Vidmar
Department of Theoretical Physics, J. Stefan Institute · Department of Physics, Faculty of Mathematics and Physics, University of Ljubljana
cond-mat.stat-mech, cond-mat.dis-nn, quant-ph
Submitted: 2026-08-03
Updated: 2026-11-18
Journal ref: Phys. Rev. B 114, 254204 (2026)
DOI: 10.1103/n93t-5lw8
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 82/100
The gist: The paper introduces a two-parameter ansatz for describing the continuous onset of Eigenstate Thermalization (ETH) violation between conventional ETH and its complete breakdown…
Key concepts
- Eigenstate Thermalization (ETH)
- A property where the energy levels of a quantum system behave statistically like those of a thermal ensemble when viewed from a large enough perspective. Conventional ETH describes how this thermal behavior emerges in systems with many degrees of freedom.
- Fading Ergodicity
- One mechanism for ETH violation where the system's ergodicity breaks down gradually as the system size increases. This is characterized by a parameter that tends toward one, indicating a continuous change in behavior.
- Trapped Ergodicity
- A distinct scenario where ETH is violated in finite systems but restored when considering the thermodynamic limit. This occurs when the ergodization length scales differently, showing a shift on the decay plot consistent with conventional ETH.
Terminology
Summary
The paper introduces a two-parameter ansatz for describing the continuous onset of Eigenstate Thermalization (ETH) violation between conventional ETH and its complete breakdown, unifying fading ergodicity and trapped ergodicity <ref:2608. p. 002744#pg2>.
Introduction to ETH Breakdown Scenarios
The authors introduce a two-parameter ansatz for ETH violation that unifies and distinguishes two mechanisms: fading ergodicity and trapped ergodicity <ref:2608.02744#pg4>. Fading ergodicity captures the established route to ergodicity breaking, while trapped ergodicity describes a distinct scenario where ETH is violated in finite systems but restored in the thermodynamic limit <ref:2608. p. 002744#pg4>. They test this framework in spin-1/2 J1-J2 chains with on-site disorder and linear potential, finding that the observed ETH violation is consistent with trapped ergodicity <ref:2608. p. 002744#pg4>.
Characterization of Conventional ETH and Violation Metrics
Conventional ETH is formulated by Srednicki as an ansatz for matrix elements where the variance of diagonal matrix element fluctuations is given by Equation (3) <ref:2608.02744#pg6>. The dominant contribution to this variance is expressed as a scaling proportional to exp[-L - ξ(h, L)/η0] <ref:2608. p. 102744#pg7>. The ergodization length ξ is defined such that only at L >> ξ does the system exhibit conventional ETH <ref:2608. p. 002744#pg8>.
The Two-Parameter Ansatz for Violation
The two-parameter ansatz for the ergodization length is introduced as ξ(h, L) = θ(h)L + λ(h) <ref:2608.02744#pg9>. Plugging this into the scaling relation yields a variance scaling proportional to exp[-(1 - θ(h))L - λ(h)/η0]. p. 1 The limit θ = 0 corresponds to conventional ETH, while the two scenarios are characterized by either theta → 1 (fading ergodicity) or λ → O(L) (trapped ergodicity). p. 2
Distinguishing Fading Ergodicity and Trapped Ergodicity
Fading ergodicity is characterized by the parameter θ(h), which increases towards 1, with a critical point h = hc where θ(h = hc) = 1 and σ2 = O(1). Trapped ergodicity is characterized by λ(h) = ξ(h), which increases with h, manifesting as a shift on the horizontal axis of the decay of σ2 consistent with conventional ETH. p. 2
Numerical Evidence in Spin Chains
Numerical studies on the disordered J1-J2 spin-1/2 chain show that for weak disorder, the decay of σ2 is consistent with conventional ETH, but at larger disorder, horizontal shifts of the decays indicate trapped ergodicity. The extracted parameters θ(h) and λ(h) are found to be roughly constant near zero for small h and increasing functions of h for trapped ergodicity. p. 7 This behavior is a clear signature of trapped ergodicity, which is very different from fading ergodicity where θ(h) increases with h. p. 3 The scaling variable L - λ∞(h) shows perfect scaling collapse in the regime when ln ˜σ2 ∝ L - λ∞(h), suggesting relevance of trapped ergodicity to describe ETH violation. p. 8
Conclusion and Significance
The main result is the introduction of a common theoretical framework to describe two distinct scenarios of the ETH breakdown. The authors show that trapped ergodicity applies to widely studied models such as the disordered J1-J2 spin-1/2 chain. p. 3 Their results offer a new perspective into finite-size effects at the boundary of ergodicity and open doors towards establishing firm connections between quantum thermalization and ergodicity breaking transitions. p. 3
Acknowledgments
The authors acknowledge support from the Slovenian Research and Innovation Agency (ARIS), Research core funding Grants No. P1-0044, N1-0273 and J1-50005, as well as the Consolidator Grant Boundary-101126364 of the European Research Council (ERC). p. 4 They gratefully acknowledge the High Performance Computing Research Infrastructure Eastern Region (HCP RIVR) consortium [81] and European High Performance Computing Joint Undertaking (EuroHPC JU) [82] for funding this research by providing computing resources of the HPC system Vega at the Institute of Information sciences [83]. p. 4 The references listed are from various sources, including Deutsch, Srednicki, Rigol, and others. p. 4
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved ETH violation classification: The AI system can distinguish between fading ergodicity, characterized by parameter theta tending towards 1, and trapped ergodicity, characterized by parameter lambda increasing with h while theta remains close to zero. This allows for a more nuanced description of how eigenstate thermalization breaks down in finite systems.
-
Predictive modeling of ergodization length: The system can predict the scaling of the variance of diagonal matrix elements using the two-parameter ansatz sigma 2 proportional to exp (-[1 - theta(h)]L - lambda(h) / eta0). This enables extrapolation to larger system sizes based on observed finite-size data.
-
Identification of ergodicity breaking regimes: The AI can identify the crossover point h star where the scaling changes from conventional ETH-like behavior to complete breakdown of ETH, which occurs when lambda(h) departs from its L independent function. This helps pinpoint critical points in parameter space related to ergodicity transitions.
-
Characterization of observable behavior: The system can analyze the rescaled variance ln sigma 2 plus theta0L / eta0 against the scaling variable L minus lambda infinity(h). This allows for a quantitative test of the trapped ergodicity hypothesis in specific models like the disordered J1-J2 chain.
-
Parameter extraction from numerical data: The AI can extract parameters theta(h) and lambda(h) by fitting linear functions to ln sigma 2 versus L, using equality instead of proportionality as described in Eq. (8). This provides a systematic method for quantifying the nature of ETH violation in simulation results.
Sources
- Weak eigenstate thermalization with large deviation bound
- Computing eigenpairs of quantum many-body systems with Polfed.jl
Related papers
- Emergent long-tail dynamics in driven magnets with dynamical frustration
- Statistical Mechanics of Dynamical System Identification
- Second law of thermodynamics in closed quantum many-body systems
- Chemical potentials from structure factors: II. Charged multi-component mixtures
- Genuine quantum scars in Floquet chaotic many-body systems
- Stochastic Thermodynamics for Autoregressive Generative Models: A Non-Markovian Perspective